Anthropic tells US senators and the White House that Alibaba's Qwen lab ran the largest known distillation campaign on Claude (28.8M exchanges)
Bloomberg reported on June 24, 2026 that Anthropic had sent a letter (dated June 10) to the Senate Banking Committee and White House officials accusing operators linked to Alibaba's Qwen lab of "adversarial distillation": about 28.8 million exchanges with Claude through nearly 25,000 fraudulent accounts between April 22 and June 5, 2026, targeting coding and agentic reasoning. Anthropic called it the largest extraction campaign it had seen, bigger than the DeepSeek, Moonshot and MiniMax campaigns it disclosed in February combined.
Key facts
- Letter dated June 10, 2026, to Senate Banking Committee chair Tim Scott and ranking member Elizabeth Warren, also shared with White House officials; first reported by Maggie Eastland (Bloomberg) on June 24
- Scale: ~28.8 million exchanges via almost 25,000 fraudulent accounts over 44 days (April 22–June 5, 2026), aimed at software engineering and agentic reasoning
- For comparison, Anthropic's Feb 23 report counted 16M+ exchanges via ~24,000 accounts across DeepSeek, Moonshot and MiniMax
- Anthropic declined to discuss specifics but stressed coordinated action between government and industry; Alibaba did not comment (TNW). Alibaba's US shares fell nearly 3% (Yahoo Finance/Stocktwits)
- Response in Congress: Senators Bill Hagerty and Andy Kim planned an NDAA amendment directing Commerce to blacklist Chinese firms found doing this; a bipartisan House bill by Reps. Bill Huizenga and Sydney Kamlager-Dove was also under consideration (TNW)
- Context (TNW): the White House flagged distillation as a national-security concern in April 2026 (OSTP director Michael Kratsios memo), and the Pentagon added Alibaba to its Chinese military companies list on June 8, 2026
- Grey market: Chinese 'transfer stations' (中转站) resell Claude API access at roughly 10% of list price via fake trial accounts, stolen cards and shared Max plans, sometimes substituting cheaper models, and resell users' prompts and outputs as training data (ChinaTalk researcher Zilan Chen, May 2026, via AI Times; Tom's Hardware). Chosun Ilbo and ZDNet Korea revisited the black market on Oct 3–4, 2026
- Related controversy (AI Times, July 2): Claude Code had run an experiment since March 2026 that flagged likely Chinese users and links to Chinese AI labs; after it surfaced on Reddit on June 30, Anthropic said it was 'meant to prevent account abuse from unauthorized resellers and protect against distillation' and removed it (not checked against a primary source)
- Re-circulated Oct 3–4, 2026 by Stocktwits ('Anthropic Writes To White House Accusing Alibaba...'), which is why it appeared in that day's sweep; no new letter was found
What happened
Anthropic did not publish the letter. Its contents are known from Bloomberg's report and follow-ups (CNBC, TNW, Tech Times, Japan Times). According to them, Anthropic said Qwen-linked operators used thousands of fake accounts and proxies to get around its ban on access from China, and asked lawmakers for stronger enforcement against foreign labs that copy US models' outputs.
Why it matters
It moved the distillation dispute from blog posts to Congress and put Alibaba, China's biggest open-weights lab, at its centre. It also fed the September threat report, which named seven China-based labs, and China's own CAC probe of DeepSeek and Moonshot.
Unverified: the letter itself was not seen. Figures come from Bloomberg's sourcing as relayed by other outlets (Bloomberg and CNBC are paywalled or blocked here). The Claude Code tracking detail comes from Korean press and was not checked against Anthropic or The Information.
Changelog
- 2026-10-04: created (sweep 2026-10-04: Stocktwits re-run of the June story; Chosun Ilbo black-market feature)
People
Related events
- Anthropic accuses DeepSeek, Moonshot and MiniMax of 'distillation attacks': 16M+ Claude exchanges via ~24,000 fraudulent accounts ★★★
- Anthropic threat intelligence report: AI-orchestrated cyberattacks and distillation by Chinese labs ★★★
- China's cyberspace regulator probes DeepSeek and Moonshot over possible data leaks to Anthropic via Claude ★★★
- Alibaba launches Qwen3.8-Max (2.4T MoE) and open-sources the Qwen3.8 family ★★★★
Sources (10)
- pressBloomberg: Anthropic accuses Alibaba of 'illicitly' accessing its AI models (June 24)
- pressCNBC: Anthropic accuses Alibaba of campaign to 'brazenly' and 'illicitly' extract AI capabilities
- pressTNW: Anthropic accuses Alibaba of running largest distillation campaign against Claude
- pressTech Times: Alibaba ran largest known AI theft campaign against Claude, Anthropic tells Senate
- pressJapan Times: Anthropic accuses Alibaba of 'illicitly' accessing AI models
- pressYahoo Finance (Stocktwits): Anthropic accuses Alibaba of illicit access to Claude AI models
- pressTom's Hardware: Chinese grey market sells Claude API access at 90% off
- pressAI Times (Korean): Claude sold at 10% of price on China's black market (May 10)
- pressAI Times (Korean): Claude Code's secret tracking of Chinese users withdrawn (July 2)
- pressZDNet Korea: China's token grey market sells US AI at 10% of list price (Oct 3)
id: 2026-06-24-anthropic-accuses-alibaba-qwen-distillation · updated 2026-10-04 · open in the interactive timeline